Linux-BTRFS Archive on lore.kernel.org
 help / color / Atom feed
From: Qu Wenruo <quwenruo.btrfs@gmx.com>
To: Marc Lehmann <schmorp@schmorp.de>, linux-btrfs@vger.kernel.org
Subject: Re: first mount(s) after unclean shutdown always fail
Date: Wed, 1 Jul 2020 09:30:25 +0800
Message-ID: <36fcfc97-ce4c-cce8-ee96-b723a1b68ec7@gmx.com> (raw)
In-Reply-To: <20200701005116.GA5478@schmorp.de>

[-- Attachment #1.1: Type: text/plain, Size: 5375 bytes --]



On 2020/7/1 上午8:51, Marc Lehmann wrote:
> Hi!
> 
> I have a server with multiple btrfs filesystems and some moderate-sized
> dmcache caches (a few million blocks/100s of GBs).
> 
> When the server has an unclean shutdown, dmcache treats all cached blocks
> as dirty. This has the effect of extremely slow I/O, as dmcache basically
> caches a lot of random I/O, and writing these blocks back to the rotating
> disk backing store can take hours. This, I think, is related to the
> problem.
> 
> When the server is in this condition, then all btrfs filesystems on slow
> stores (regardless of whether they use dmcache or not) fail their first
> mount attempt(s) like this:
> 
>    [  173.243117] BTRFS info (device dm-7): has skinny extents
>    [  864.982108] BTRFS error (device dm-7): open_ctree failed
> 
> Recent kernels sometimes additionally fail like this (super_total_bytes):
> 
>    [  867.721885] BTRFS info (device dm-7): turning on sync discard
>    [  867.722341] BTRFS info (device dm-7): disk space caching is enabled
>    [  867.722691] BTRFS info (device dm-7): has skinny extents
>    [  871.257020] BTRFS error (device dm-7): super_total_bytes 858976681984 mismatch with fs_devices total_rw_bytes 1717953363968
>    [  871.257487] BTRFS error (device dm-7): failed to read chunk tree: -22
>    [  871.269989] BTRFS error (device dm-7): open_ctree failed

This looks like an old fs with some bad accounting numbers.

Have you tried btrfs rescue fix-device-size?

Thanks,
Qu
> 
> all the filesystems in question are mounted twice during normal boots,
> with diferent subvolumes, and systemd parallelises these mounts. This might
> play a role in these failures.
> 
> Simply trying to mount the filesystems again then (usually) succeeds with
> seemingly no issues, so these are spurious mount failures. These repeated
> mount attewmpts are also much faster, presumably because a lot of the data
> is already in memory.
> 
> As far as I am concerned, this is 100% reproducible (i.e. it happens on every
> unclean shutdown). It also happens on "old" (4.19 era) filesystems as well as
> on filesystems that have never seen anything older than 5.4 kernels.
> 
> It does _not_ happen with filesystems on SSDs, regardless of whether they
> are mounted multiple times or not. It does happen to all filesystems that
> are on rotating disks affected by dm-cache writes, regardless of whether
> the filesystem itself uses dmcache or not.
> 
> The system in question is currently running 5.6.17, but the same thing
> happens with 5.4 and 5.2 kernels, and it might have happened with much
> earlier kernels as well, but I didn't have time to report this (as I
> secretly hoped newer kernels would fix this, and unclean shutdowns are
> rare).
> 
> Example btrfs kernel messages for one such unclean boot. This involved
> normal boot, followed by unsuccessfull "mount -va" in the emergency shell
> (i.e. a second mount fasilure for the same filesystem), followed by a
> successfull "mount -va" in the shell.
> 
> [  122.856787] BTRFS: device label LOCALVOL devid 1 transid 152865 /dev/mapper/cryptlocalvol scanned by btrfs (727)
> [  173.242545] BTRFS info (device dm-7): disk space caching is enabled
> [  173.243117] BTRFS info (device dm-7): has skinny extents
> [  363.573875] INFO: task mount:1103 blocked for more than 120 seconds.
> the above message repeats multiple times, backtrace &c has been removed for clarity
> [  484.405875] INFO: task mount:1103 blocked for more than 241 seconds.
> [  605.237859] INFO: task mount:1103 blocked for more than 362 seconds.
> [  605.252478] INFO: task mount:1211 blocked for more than 120 seconds.
> [  726.069900] INFO: task mount:1103 blocked for more than 483 seconds.
> [  726.084415] INFO: task mount:1211 blocked for more than 241 seconds.
> [  846.901874] INFO: task mount:1103 blocked for more than 604 seconds.
> [  846.916431] INFO: task mount:1211 blocked for more than 362 seconds.
> [  864.982108] BTRFS error (device dm-7): open_ctree failed
> [  867.551400] BTRFS info (device dm-7): turning on sync discard
> [  867.551875] BTRFS info (device dm-7): disk space caching is enabled
> [  867.552242] BTRFS info (device dm-7): has skinny extents
> [  867.565896] BTRFS error (device dm-7): open_ctree failed
> [  867.721885] BTRFS info (device dm-7): turning on sync discard
> [  867.722341] BTRFS info (device dm-7): disk space caching is enabled
> [  867.722691] BTRFS info (device dm-7): has skinny extents
> [  871.257020] BTRFS error (device dm-7): super_total_bytes 858976681984 mismatch with fs_devices total_rw_bytes 1717953363968
> [  871.257487] BTRFS error (device dm-7): failed to read chunk tree: -22
> [  871.269989] BTRFS error (device dm-7): open_ctree failed
> [  872.535935] BTRFS info (device dm-7): disk space caching is enabled
> [  872.536438] BTRFS info (device dm-7): has skinny extents
> 
> Example fstab entries for the mounts above:
> 
> /dev/mapper/cryptlocalvol       /localvol       btrfs           defaults,nossd,discard                  0       0
> /dev/mapper/cryptlocalvol       /cryptlocalvol  btrfs           defaults,nossd,subvol=/                 0       0
> 
> I don't need assistance, I merely write this in the hope of btrfs being
> improved by this information.
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

  reply index

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-01  0:51 Marc Lehmann
2020-07-01  1:30 ` Qu Wenruo [this message]
2020-07-01 20:14   ` Marc Lehmann
2020-07-01 23:45     ` Qu Wenruo
2020-07-01 23:55       ` Marc Lehmann
2020-07-02  0:02         ` Qu Wenruo
2020-07-02  1:11           ` Marc Lehmann
2020-07-02  1:28             ` Qu Wenruo
2020-07-02  2:13               ` Marc Lehmann
2020-07-02 18:31 ` Zygo Blaxell
2020-07-03  8:04   ` Marc Lehmann

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=36fcfc97-ce4c-cce8-ee96-b723a1b68ec7@gmx.com \
    --to=quwenruo.btrfs@gmx.com \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=schmorp@schmorp.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-BTRFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-btrfs/0 linux-btrfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-btrfs linux-btrfs/ https://lore.kernel.org/linux-btrfs \
		linux-btrfs@vger.kernel.org
	public-inbox-index linux-btrfs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-btrfs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git